This approach would require a quick access to the tactical results, and would slow down the random games. The pattern database should be built a priori and should not introduce too much bias into the random games. Bruno Bouzy There might be more efficient ways to analyze a random game and decide whether the value of a move is the same as if it was played at the root. We have set up experiments to assess ideas such as progressive pruning, transpositions, temperature, simulated annealing and depth-two tree search within the Monte Carlo framework.

Author:Zulugore Guzragore
Language:English (Spanish)
Published (Last):15 August 2010
PDF File Size:19.5 Mb
ePub File Size:10.58 Mb
Price:Free* [*Free Regsitration Required]

Olga and Oleg We have developed two go programs based on the basic idea above: The value of those subgoals could, then, be evaluated by linking them to the final score. We believe that the statistical search is an alternative to tree search [Junghanns, ]worth considering in practice.

Indigo and GnuGo Indigo [Bouzy, ]is a classical program using tree search and knowl- edge. This progressive pruning algorithm is similar to the one described in [Billings et al. Furthermore, two-level hedging algorithms are more effective than one-level hedging algorithms, and three levels are not better than two levels.

How do the uses of transpositions and progressive pruning compare in strength? The amount of random put in the games was controlled with the temperature; it was set high at the begin- ning and gradually decreased. The author showed that the expected outcome is a powerful heuristic. When compared to knowledge based ap- proaches, this approach is very easy to implement but its weakness lies in the tactics.

Two of them were already present in Gobble, namely transpositions which means making statistics not only for the first move but for all moves of the random games and simulated annealing. By clicking accept or continuing to use the site, you agree to the terms outlined in our Privacy PolicyTerms of Serviceand Dataset License. The pattern database should be built a priori and should not introduce too much bias into the random games. In figure 2 the values of the moves A for black and B for white largely depend on the order in which they are played.

The minimal number of random games without pruning is set to Let us start the random games from the root by two given moves, one move for the friendly side, and, then, one move for the opponent, and make statistics on the terminal position evaluation for each node situated at depth 2 in the min-max tree.

Then, he evaluated a move by computing the average of the scores of the random games where it had been played. Game-tree searching by min-max approximation. First, this has been estimated with a match against two versions of Oleg.

This section deals with two previous works [Abramson, ]and [Bruegmann, ]. The goal of this paper is to present Computer Go by showing the links between existing studies on Computer Go and different AI related domains: This sole domain- dependent knowledge in Gobble is necessary to ensure that the random games actually finish.

A review of game-tree pruning. Besides, Olga burno a fuzzy and optimistic definition: Therefore, brujo number of candidate moves decreases while the process is running. Their results highlight the … More. Section 3 focusses on the main ideas underlying our work. The former is the basic idea, the latter is what was performed in Gobble. Probably the best moves are played early and thus, get a more accurate evaluation.

It costs nothing in execution time but the move generator remains incomplete and always contains errors. It does not necessarily need domain-dependent knowledge but its cost is exponential in the depth search. In the latter, two lists of moves were maintained for both players, and the moves in the random games were played in the order of the lists if the move in the list is not legal, we just take the next in the list.

Besides, we have tried to enhance our programs with a depth-two tree search which did not work well. Click here to sign up. In Progressive Pruning PPafter a minimal number of random games per movea move is pruned as soon as it is statistically inferior to another move.

However, optimizing the program very roughly is important. This is an adaptation of [Abramson, ]. It might be possible to link the value of a move to more local subgoals from which we could establish statistics. Artificial Intelligencepages 39— On one hand, mathematical morphology is a very powerful tool within bbruno processing community.

Nevertheless, this approach looks promising. Developments On Monte Carlo Go. In incomplete information games, such as poker [Billings et al. Related Articles.


Author: Bruno Bouzy



Bruno Paillard N.P.U. 2004


Related Articles